STARD 2015 Checklist: 30 Items for Diagnostic Accuracy
The 30 STARD 2015 items for diagnostic accuracy studies, what reviewers object to in each block, and a checker for your manuscript.
STARD 2015 is the guideline for reporting a diagnostic accuracy study: one that compares a test (the index test) with a reference standard and reports sensitivity, specificity, predictive values or the area under the curve. It has 30 items and requires a participant flow diagram. It replaces STARD 2003, which had 25.
By Víctor Ciudad-Fernández, PhD · Updated: October 6, 2026 · Source: Bossuyt et al. (2015)
What STARD 2015 is and when to use it
STARD (Standards for Reporting of Diagnostic Accuracy Studies) brings together what the article has to contain so that readers can judge whether the published accuracy is credible and for which patients it holds. The 2015 version is by Bossuyt and colleagues, with a 2016 explanation and elaboration paper that gives the reasoning for each item.
A test’s accuracy depends heavily on who it is done on and how it is read. That is why STARD insists on things that go unnoticed in other designs: how participants were recruited, whether whoever read one test knew the result of the other, and what was done with indeterminate results.
Studies that compare a test (laboratory, imaging, a clinical sign, a screening questionnaire) with a reference standard and report accuracy measures.
Validating a cut-off on a scale against a clinical diagnosis (for example, against a structured interview): that is diagnostic accuracy too.
If the study develops or validates a prediction model that combines several variables into a risk score, the guideline is TRIPOD+AI, not STARD.
To assess the risk of bias of diagnostic studies within a systematic review, use QUADAS-2, which is not a reporting guideline.
The 30 items, section by section
Some items have sub-items (10a and 10b, 12a and 12b, 13a and 13b, 21a and 21b). What follows explains each one in plain words; for submission, use the official checklist. Under each section you have what reviewers usually object to in it.
Title and abstract
1. The title or abstract identifies it as a diagnostic accuracy study, with at least one accuracy measure (sensitivity, specificity, predictive values or area under the curve).
2. A structured abstract with design, methods, results and conclusions.
What reviewers usually object to:
The title says "validation" without making clear it is a diagnostic accuracy study, and the editor does not know which checklist to review it against.
The abstract gives sensitivity and specificity without their confidence intervals.
Introduction
3. The scientific and clinical background, including the intended use of the index test and its role in the clinical pathway.
4. The objectives and hypotheses.
What reviewers usually object to:
It does not say where the test would sit in the clinical pathway (screening, triage, replacement, add-on), and without that it is impossible to judge whether the numbers are useful.
Methods
5. Whether data collection was planned before the tests were done (prospective) or after (retrospective).
6. The eligibility criteria.
7. What the identification of potentially eligible participants was based on (symptoms, results of previous tests, inclusion in a registry).
8. Where and when potentially eligible participants were identified: setting, location and dates.
9. Whether participants formed a consecutive, random or convenience series. (checked by the checker)
10a. The index test, in enough detail to replicate it.
10b. The reference standard, in enough detail to replicate it. (checked by the checker)
11. Why that reference standard was chosen, if there were alternatives.
12a. The definition of and rationale for the index test’s cut-offs or result categories, separating prespecified from exploratory ones. (checked by the checker)
12b. The same for the reference standard.
13a. Whether the people performing or reading the index test knew the clinical information and the reference standard result. (checked by the checker)
13b. Whether the people assessing the reference standard knew the clinical information and the index test result. (checked by the checker)
14. The methods for estimating or comparing measures of diagnostic accuracy.
15. How indeterminate results of the index test or reference standard were handled. (checked by the checker)
16. How missing data on the index test and reference standard were handled.
17. Any analyses of variability in accuracy, separating planned from exploratory ones.
18. The intended sample size and how it was determined. (checked by the checker)
What reviewers usually object to:
The sample is a convenience or case-control type sample (severe cases versus healthy people): accuracy comes out inflated by spectrum bias.
Not every participant received the reference standard, or different standards were used depending on the index test result (verification bias).
The cut-off was chosen in the same sample with the Youden index and is presented as if it had been prespecified.
There is no statement that whoever read the index test was unaware of the reference standard result.
Indeterminate results were excluded without saying how many there were.
Results
19. The flow of participants, with a diagram. (checked by the checker)
20. The baseline demographic and clinical characteristics of participants.
21a. The distribution of disease severity in those with the target condition.
21b. The distribution of alternative diagnoses in those without it.
22. The time interval between the index test and the reference standard, and any clinical interventions in between.
23. The cross-tabulation of index test results against reference standard results (or their distribution, if the test is continuous). (checked by the checker)
24. The estimates of diagnostic accuracy with their precision (for example, 95% confidence intervals).
25. Any adverse events from performing the index test or the reference standard.
What reviewers usually object to:
The 2×2 table is missing: the reviewer cannot recompute sensitivity, specificity or predictive values.
Predictive values are presented as transferable to other settings, when they depend on the prevalence in this sample.
There is no flow diagram, or the one there does not match the N analyzed.
Discussion
26. The study limitations: sources of bias, statistical uncertainty and generalizability.
27. The implications for practice, with the intended use and clinical role of the index test.
What reviewers usually object to:
The test is recommended for a use (screening the general population) different from the one studied (patients referred to a clinic).
Other information
28. The registration number and the name of the registry.
29. Where the full protocol can be accessed.
30. The sources of funding and other support, and the role of the funders.
What reviewers usually object to:
The study was not registered and has no protocol, so there is no way to know whether the cut-off and the analysis were planned.
Check your manuscript against STARD 2015
Paste the article (Methods and Results at a minimum). The checker searches the text for 8 of the 30 STARD 2015 items, the ones a text search can decide without reading (participant flow diagram, reference standard described, blinding of test readers, consecutive, random or convenience series, test positivity cut-off, index test × reference standard cross-tabulation, indeterminate results, intended sample size), and shows you the sentence that supports each one. It runs in your browser: nothing is uploaded or stored.
The Professional Review in Previum does not check STARD yet: these items are only checked on this page.
Checklist checker: Detects the design and picks the guideline for you.
Previum · AI peer review. A checklist tells you whether something appears in the text; it does not tell you whether it was done right. Once the manuscript is complete, see what Reviewer 2 would object to before you submit: findings anchored to your text, in the margin. Review my manuscript
What changed from STARD 2003
STARD 2003 had 25 items; the 2015 version has 30. The new items include the structured abstract, the intended use and clinical role of the test, the hypotheses, the sample size, registration, access to the protocol and funding.
The participant flow diagram, previously recommended, is now required in every report.
How to cite the guideline
If your Methods section says you followed STARD 2015, cite the statement (and the explanation and elaboration if you rely on it), not this page.
Bossuyt, P. M., Reitsma, J. B., Bruns, D. E., Gatsonis, C. A., Glasziou, P. P., Irwig, L., Lijmer, J. G., Moher, D., Rennie, D., de Vet, H. C. W., Kressel, H. Y., Rifai, N., Golub, R. M., Altman, D. G., Hooft, L., Korevaar, D. A., & Cohen, J. F. (2015). STARD 2015: An updated list of essential items for reporting diagnostic accuracy studies. BMJ, 351, h5527. https://doi.org/10.1136/bmj.h5527
Cohen, J. F., Korevaar, D. A., Altman, D. G., Bruns, D. E., Gatsonis, C. A., Hooft, L., Irwig, L., Levine, D., Reitsma, J. B., de Vet, H. C. W., & Bossuyt, P. M. M. (2016). STARD 2015 guidelines for reporting diagnostic accuracy studies: Explanation and elaboration. BMJ Open, 6(11), e012799. https://doi.org/10.1136/bmjopen-2016-012799
Thirty, some with sub-items. The previous version, STARD 2003, had 25.
When do I use STARD and when TRIPOD?
STARD when you evaluate how well a test performs against a reference standard. TRIPOD (now TRIPOD+AI) when you develop or validate a model that combines several variables to predict a diagnosis or a prognosis.
Can I use STARD to validate a screening questionnaire?
If you compare the questionnaire score with a reference diagnosis and report sensitivity and specificity at a cut-off, yes: it is a diagnostic accuracy study. If what you validate is the factor structure or the reliability, no; there the reference framework is COSMIN.
Is the flow diagram mandatory?
In STARD 2015, yes: item 19 asks you to describe the flow of participants with a diagram, from those eligible to those with results on both tests.